Re: "Case definition in case-control studies of the efficacy of screening".

Evaluation of the efficacy of screening programs involves the comparison of morbidity or mortality rates between screened and unscreened populations. Such comparisons are made most convincingly by means of experiments (1, 2), exemplified by the study of screening for breast cancer carried out by the Health Insurance Plan of Greater New York (HIP) (3). However, there are serious practical difficulties with experimental studies of screening: they are expensive, may require many years to complete, and usually can be done only when new screening procedures are introduced. Once a procedure is established as a part of medical practice, as the Papanicolaou (Pap) smear is, it may be necessary to rely on nonexperimental methods to evaluate it. Nonexperimental studies are more susceptible than are experimental ones to "confounding." For example, social class (as reflected in education or income) is directly related to frequency of screening with the Pap smear, but is inversely related to cervical cancer rates; relatively low rates of cervical cancer among frequently screened women could be secondary to their relatively high social class. Despite this drawback of the nonexperimental approach, there may be times when it will be able to provide information that cannot be obtained otherwise. (A discussion of the roles of experi-